Autonomous systems working in concert — and with humans, too
Autonomous and connected systems (ACS) are groupings of autonomous systems — like robots and vehicles — that interact with one another via communication and sensing networks. By working as a cohesive whole, ACS can provide more autonomy than individual systems and accomplish sophisticated tasks beyond individual systems’ capabilities. Examples include a drone swarm exploring a large area, a platoon of driverless trucks on a highway, or a team of autonomous ground robots and aerial vehicles collaborating in a rescue operation.
The trick is getting this network of autonomous systems to work as a cohesive whole. We recently launched the Center for Innovation in Control, Optimization, and Networks (ICON) to help researchers and practitioners integrate diverse multidisciplinary techniques and ideas to enable the next generation of ACS. I co-direct the center with Shreyas Sundaram, associate professor of electrical and computer engineering. It currently brings together nearly 50 researchers from across the College of Engineering.
One of the many initiatives by ICON researchers includes a multi-university project to develop AI-assisted adaptive planning for multi-agent “human-in-the-loop” systems, which share autonomy with people. Researchers are creating techniques for environmental perception and situational awareness, as well as decision-making and task assignment in changing environments. We’re studying real-time distributed solutions to control assets, and using machine learning to continuously improve performance.
ICON researchers also are investigating methods to include human inputs and corrections to achieve mixed human-machine autonomy, as people’s interaction with autonomous systems becomes ubiquitous in consumer products, transportation systems, manufacturing, and other domains. We are exploring how to design responsive, personalized systems that provide a high confidence of reliability, and developing a human-centric architecture for “cognitive autonomy,” which couples human psychophysiological and behavioral measures with performance metrics. This novel architecture, enabled by advances in computation, communication and control, leverages the strengths of both humans and automation to achieve new levels of performance and safety.
Cyber-physical systems that adapt to the human — and can account for a person’s ongoing adaptation to the system — could have enormous impacts in everyday life, as well as in specialized domains like biomedical devices and systems, transportation systems, manufacturing, and military applications. They could significantly reduce training time, increasing human familiarity with the system before operation in safety-critical environments. AI and machine learning can assist the system in many ways, such as in object recognition and human-machine communication. For example, in a search-and-rescue scenario, an AI-aided system can allow a human commander to offer input on mission parameters while drones provide feedback and even suggestions in synthesized natural language. For these complex situations, we still need to involve humans in a mixed autonomy model.
ICON’s goal is to address fundamental challenges in these and other emerging research areas and create innovative solutions to grand challenges. We also will offer customizable curricula and instructional delivery approaches to satisfy educational needs, as well as collaborate with industries, governmental agencies, and national labs to tackle national and global priorities.
Our hope is that future autonomous and connected systems will be able to work with and beside human operators independently or with little human guidance, especially in missions involving heavy labor or dangerous situations.
Shaoshuai Mou, PhD
School of Aeronautics and Astronautics
College of Engineering, Purdue University